Open Issues Need Help
View All on GitHub Docs still document -emit=jit, and one-block-one-thread kernels about 1 hour ago
documentation good first issue
An MLIR-based Compiler Infrastructure for Neural Networks
C++
#deep-learning#llvm#machine-learning-compiler#mlir#mlp#neural-networks#transformer
chooseBlockSize can silently produce a one-thread block about 2 hours ago
enhancement good first issue
An MLIR-based Compiler Infrastructure for Neural Networks
C++
#deep-learning#llvm#machine-learning-compiler#mlir#mlp#neural-networks#transformer
hextir.for accepts any kind string, and silently ignores "vectorized" and "unrolled" about 2 hours ago
bug good first issue
An MLIR-based Compiler Infrastructure for Neural Networks
C++
#deep-learning#llvm#machine-learning-compiler#mlir#mlp#neural-networks#transformer
CUDA HAL reloads the device image on every dispatch about 2 hours ago
enhancement good first issue
An MLIR-based Compiler Infrastructure for Neural Networks
C++
#deep-learning#llvm#machine-learning-compiler#mlir#mlp#neural-networks#transformer
Wire up the affine pass, or stop accepting -emit=mlir-affine about 2 hours ago
bug good first issue
An MLIR-based Compiler Infrastructure for Neural Networks
C++
#deep-learning#llvm#machine-learning-compiler#mlir#mlp#neural-networks#transformer